Novel Class Discovery in Chest X-rays via Paired Images and Text

Authors: Jiaying Zhou, Yang Liu, Qingchao Chen

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on eight subset splits of MIMIC-CXRJPG dataset show that our method improves the clustering performance of unlabeled classes by about 10% on average compared to state-of-the-art methods.
Researcher Affiliation Academia Jiaying Zhou1, 2, Yang Liu3, Qingchao Chen1, 2, 4 * 1National Institute of Health Data Science, Peking University, Beijing, China 2Institute of Medical Technology, Peking University Health Science Center, Beijing, China 3Wangxuan Institute of Computer Technology, Peking University, Beijing, China 4 National Key Laboratory of General Artificial Intelligence, Beijing, China
Pseudocode No The paper does not contain a pseudocode block or an explicitly labeled algorithm section.
Open Source Code Yes Code is available at: https://github.com/zzzzzzzzjy/MMNCD-main.
Open Datasets Yes MIMIC-CXR-JPG Dataset(Johnson et al. 2019b) This dataset contains 377,110 chest X-ray images from 65,379 patients. Each image is provided with 14 labels derived from two natural language processing tools applied to the corresponding free-text radiology reports.
Dataset Splits No The paper mentions 'Validation set accuracy' in Figure 1 (b) but does not provide explicit details on the size or split methodology for a validation dataset, such as percentages or sample counts.
Hardware Specification No The paper does not provide specific details regarding the hardware used for experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper mentions the use of Res Net-50 and Bio Clinical BERT but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, or other libraries).
Experiment Setup Yes We train our model in two stages... Then we conduct novel class discovery on our network with 200 epochs. All experiments are conducted with a fixed batch size of 128.